Change detection in urban built-up volume using deep learning based segmentation techniques

Mapping cities is difficult due to the dense and varied development of infrastructures in urban contexts. To adapt cities for modern faculties, regular monitoring of developing infrastructures is essential. The built-up land use map derived from remote sensing images shows how cities expand horizontally across the landscape as they grow over decades. Recent studies have shown that when planning for a more sustainable future in metropolitan settings, adding vertical components in the computation of built-up is beneficial for better understanding. As illustrated in this paper, the built-up area of an urban region is extracted from high-resolution LISS IV satellite sensor image data using deep learning techniques. The height of an urban built-up is estimated from Cartosat-1 stereo images using photogrammetric methods and automatic terrain extraction. This study combines the use of a deep learning model with the use of high-resolution remote sensing data to extract the urban built up volume. The results demonstrate that the western and northwestern areas of the study region see greater changes in urban volume compared to the central business district and other parts of the city. This work demonstrates a step-by-step strategy for studying urban growth patterns. It helps in understanding urban growth, population mobility patterns, societal typology, environmental indicators. The outcome of such study could be useful in the evaluation of various indicators of growth such as economic growth, traffic density conditions, utility planning, policy framework such as master plan preparation, and so on for governmental authorities.